Oscillatory responses to highly predictable words differentiate between expectations based on semantic or associative contextual constraints

Monsalve, I. 1 , Pérez, A. 1 & Molinaro, N. 1, 2

1 Basque Center on Cognition, Brain and Language (BCBL)
2 Ikerbasque, Basque Foundation for Science

During language comprehension, semantic contextual information is used to generate expectations about upcoming items. This has been frequently studied through the N400 event-related potential (ERP), as a measure of facilitated lexical retrieval. However, the associative relationships in multi-word expressions (MWE) may enable the generation of categorical expectations, leading to lexical retrieval before target word onset. Processing of the upcoming word would thus involve a target-identification mechanism, possibly indexed by a P3 component. However, given their time overlap (200-500 ms post-stimulus onset), differentiating between N400/P3 ERP responses is problematic. In the present study, we analyzed EEG data from a previous experiment, which compared ERP responses to highly expected words placed either in a MWE or a regular compositional context. We focused on time-frequency and single-trial statistical analyses, in order to use individual item variability to dissociate between conditions. A significant correlation between word position and power in the theta band (7-9 Hz) was found only for MWE, providing evidence for the presence qualitative differences between conditions. Power levels within this band were lower for MWE than compositional contexts, suggesting that in the former lexical retrieval had taken place before word onset. On the contrary, gamma-power was modulated by predictability of the item in all conditions, which could reflect similar 'matching' sub-steps in the binding process between an expected representation and the external input.